Cargando…

Gradient Decomposition Methods for Training Neural Networks With Non-ideal Synaptic Devices

While promising for high-capacity machine learning accelerators, memristor devices have non-idealities that prevent software-equivalent accuracies when used for online training. This work uses a combination of Mini-Batch Gradient Descent (MBGD) to average gradients, stochastic rounding to avoid vani...

Descripción completa

Detalles Bibliográficos
Autores principales: Zhao, Junyun, Huang, Siyuan, Yousuf, Osama, Gao, Yutong, Hoskins, Brian D., Adam, Gina C.
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Frontiers Media S.A. 2021
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8645649/
https://www.ncbi.nlm.nih.gov/pubmed/34880721
http://dx.doi.org/10.3389/fnins.2021.749811
Descripción
Sumario:While promising for high-capacity machine learning accelerators, memristor devices have non-idealities that prevent software-equivalent accuracies when used for online training. This work uses a combination of Mini-Batch Gradient Descent (MBGD) to average gradients, stochastic rounding to avoid vanishing weight updates, and decomposition methods to keep the memory overhead low during mini-batch training. Since the weight update has to be transferred to the memristor matrices efficiently, we also investigate the impact of reconstructing the gradient matrixes both internally (rank-seq) and externally (rank-sum) to the memristor array. Our results show that streaming batch principal component analysis (streaming batch PCA) and non-negative matrix factorization (NMF) decomposition algorithms can achieve near MBGD accuracy in a memristor-based multi-layer perceptron trained on the MNIST (Modified National Institute of Standards and Technology) database with only 3 to 10 ranks at significant memory savings. Moreover, NMF rank-seq outperforms streaming batch PCA rank-seq at low-ranks making it more suitable for hardware implementation in future memristor-based accelerators.